This feels like the sort of architecture that starts clean and then gradually grows most of the things a workflow-native system already has. I've seen systems like this, seen companies that are built out of this idea, and built small systems like this over time.
Once you need retries, backoff, timeouts, cancellation, versioning, visibility, task routing, rate limits, leases, heartbeats, stuck-worker detection, replay/debugging semantics, workflow migration, fanout/fanin, long timers, audit trails, and operator tooling, the “just use a database” story becomes “build a poor copy of a workflow engine plus a bunch of workers.” pretty quick.
That may still be a good tradeoff for many applications, especially if Postgres is already the core operational dependency. But the comparison shouldn’t be “database vs overcomplicated orchestrator.” It’s more like “what complexity do you want to own, and what do you want to buy / offload to a professional system?”
Yeah, we've observed that too: people start implementing their own retry logic, idempotency, etc. But then they grow a hard to maintain, complex stack that's not their core business logic. There's a reason why there is a dedicated team building DBOS, every day. Because it's not that easy to build a solid durable workflows engine on Postgres.
The SKIP LOCKED pattern is fine until the worker count climbs. Then vacuum can't keep up. Dead tuples pile up, visibility map turns to swiss cheese. Queue table is tiny on disk but the planner thinks it's huge and stops using the index. It gets ugly fast.
If you look at the 95% CI on https://marginlab.ai/trackers/codex/ with N=50, it's still pretty huge (+/- 13-14% usually). I suspect it would be difficult to reasonably get a measure that numerically assesses whether an AGENTS.md is good. What you can observe though is whether the model paid attention to certain rules while editing. I.e. did the behavior you're steering away or towards take place.
The hardest thing I think is judging whether your AGENTS.md is still good based on each model release. OpenAI does release prompting guidance however to help this (and have added a skills to apply this to your prompts IIRC)
Yes, agree that low n makes overclaiming a real risk with this sort of optimization loop. Low n results can be useful directionally but can't claim superiority without expanding the dataset. If I were running this for a shared repo with real consequences / value to improving AGENTS.md, instead of just as an experiment, I would expand n by a few factors for training / holdout, depending on expected variation on the tasks.
I'm also noticing similar patterns with needing to update AGENTS.md / skills per model release. E.g with Opus 4.6 -> 4.7, it became much more instruction adherent, so instructions written for the prior model generation might cause unexpected behavior in the new generation. I'm also convinced that an optimal AGENTS.md for Codex is not the same file as an optimized CLAUDE.md for Claude - the model personalities and behaviors are so different that we probably need to tune the instructions differently as well.
Historically, “TUI” referred to “Text User Interface”, but language evolves and “Terminal User Interface” has been common usage in terminal-app/tooling communities for at least the last decade (probably more).
The distinction we (Ratatui) draw is that the rendering surface is a terminal: historically sometimes a physical hardware terminal (someone got Ratatui running on a Minitel a while back), more commonly today a terminal emulator/PTY environment.
That framing also better captures the kinds of constraints and capabilities Ratatui apps actually deal with: terminal escape sequences, cell-based rendering, alternate screen buffers, mouse handling, resize behavior, scrollback interaction, etc.
In addition, these apps don't only do text only these days as modern terminals support various graphics protocols (Sixel, Kitty, iTerm graphics) and other extensions that can allow for weird and wonderful things that are not just text.
Thanks! As I think about terminals in quite different contexts (credit card terminal, battery terminal, train terminal, etc.) I'm realizing that it's really just any kind of endpoint at all, so pretty much any human interface device (HID) should probably be called a terminal. But when taken to that level, using the acronym TUI (in the "terminal" sense) to mean "not a GUI/CLI" (as if big old consoles or emulators/PTYs thereof are the only type of terminals) seems questionable.
Sixels are far from modern :) They were introduced in the VT240 (special purpose version of the 220) and became mainstream on DEC terminals with the 320.
I think this belongs on hacker news (and unflagged) mainly because of who Tim Bray is. Notably co-inventor of XML, worked on a bunch of web standards etc.
Whether you agree or disagree with non-US citizens coming to America to engage in the advancement of technology, the important thing is to have discourse on the topic. That is in line with aims and goals of this site. This story is much less politics and much more about the impact of social policy on technologists.
As a non-US citizen myself but who has lived there for some time, I find that having and expressing an opinion on things like this is difficult due to the danger of such retaliation mentioned during border crossings and my daily life.
As another non-US citizen I feel similar. My profile as a maker and software dev with extensive infrastructure knowledge should be exactly what the US is looking for, but I already decided the US isn't a country I am going to police my own speech for.
First watching the tearing up of the societal contract in the US over the past decades turned this into a country that feels like it is on the brink of a collapse. No longer can regular Americans rely that working hard will give them a good or even decent life. This means intense pressures are at work with some very desperate people, bringing out the ugliest (but also sometimes the most beautiful) bits of humanity, all while you have richer folks at the top who act like how their country operates at large is none of their business as long as there is a cut in it for them. Why should I care about the US if you guys don't even care about it yourselves?
Second the US is highly unreliable. Laws, democracy, human rights, treaties, consumer rights — all are treated as negotiable, optional things that you use against your enemies and ignore when it affects your own. With enough money everything can be bent around. There are no principles at work that one could rely on and that is no foundation to build a life on unless you are literally ready to join the ranks of people acting like cartoon villains — something only people do who have no self-respect for their limited time on earth.
Yeah, I don't see myself even visiting the US in my lifetime.
> First watching the tearing up of the societal contract in the US over the past decades turned this into a country that feels like it is on the brink of a collapse.
Hehe, reminds me about a video I saw some years ago.
They asked ordinary citizens what a reasonable distribution of wealth should be. Like the share of overall wealth owned by the richest 10% (or 1% or whatever), vs wealth owned by the rest of the population. Let's call this ratio "X".
Then those citizens were asked what they thought the wealth distribution was actually like (let's call that "Y"). No surprise: people thought wealth was much more UNevenly distributed than it should be.
Then they showed what the distribution was like in reality (let's call that "Z"): way more extreme.
So, wealth distributed way more uneven than people thought, and that in turn way more uneven than what people thought it should be.
That alone explains a lot of what's wrong with the US imho: broken/corrupt politics, outsized influence of tech bros (including outside US), a health problem bankrupting people, people turning to drugs, a militarized police force, incarceration ratios, homeless epidemic, etc etc.
Especially in a highly developed country where total wealth is enough to have everybody live comfortably & care-free.
As opposed to some poor war-torn country where wealth is also unevenly distributed, but it's obvious and everybody knows it. And overall wealth is a lot less to begin with (so people being desperate isn't surprising).
US' downfall turning it into a banana republic, its erosion of democratic institutions, human rights etc, is a logical result from the above. I'm not expecting that to improve sadly :-(. Things will probably get worse & there will be blood in the streets.
I spoke to a few HNers who I know flagged this post. They mentioned that they flagged it because it goes against the HN guidelines - a holy grail of sorts - and that the post does not invite "curious discussion". When prompted, they mentioned they would much rather read about pro-Musk technological exploits than anything against the US.
Unfortunately, flagging is heavily abused here on HN. Criticism of adtech and attempts to defend privacy laws will bring out flagging in droves; almost as if some HN'ers salaries are dependent on opposing such laws!
Some sort of meta-moderation system to prevent abuse of flagging would be welcome.
Unfortunately HN is likely astroturfed just as much as any other forum, despite the otherwise seemingly substantiveness of HN compared to sites like Reddit. And unfortunately there's no way to cancel out bad faith flagging and downvoting the same way you can reply to bad faith commenting. It's more invisible and easier to abuse (once the karma thresholds have been reached, which isn't hard).
there are tons of posts on HN that don't invite "curious discussion"
this is one that would actually invite curious discussion if some people weren't clutching onto their "God Bless America (We're #1!)" pearls quite so tightly
More and more I'm finding that the most interesting conversation about America isn't coming from Americans anymore.
There was a time where I was quite interested in listening to people from America talk about their fascinating and crazy sounding country, but as time goes on I find that to be much more repetitive and not insightful.
Now I'm more interested in what people from other countries have to say about America, and I find it fascinating how Americans online find this unsettling and sometimes get snippy about it really weird ways.
Sometimes I wonder what political threads on HN would be like if Americans weren't allowed to participate in them. For the people in the crowd who take things more literally in know this isn't possible, it's just an interesting thought experiment.
Would that result in more 'creative' conversation?
Maybe for a while until new patterns/tropes/memes were built up by the users that could comment on them. Maybe the issue with the political discussions that too often the people talking about them are too close to them, too immersed in them?
> A vouch option for Flagged submissions would be appreciated.
AFAIK, upvoting a flagged submission cancels out the flagging to some extent. I don’t know the internals of how this process works. I’ve upvoted the submission in an effort to get it unflagged (it still may not get to the front page or may rapidly drop down though).
There used to be a time when it was just accepted that most of the good conferences/gatherings would be in the US, and it would either be important to go to, or be relatively straightforward to reach and attend (especially for Canadians), and nobody would think twice about it.
Now you have some very talented/consequential people just refusing to visit the country. Regardless of any qualms about the "content and logic", this should set off alarm bells for any American. Plus, this would've been a whine if he'd complained and had gone anyway. But he is not going, and has explained his decision. That makes it more of a statement.
I think the title is appropriate and is probably meant as a double entendre. The US does seem to be declining in many ways. Personally, visiting the US under the present government just feels too risky, they seem really hostile to foreigners.
> Honestly, despite all the hype around Rust in the community, the fact that AI can't handle lifetimes reliably makes me reluctant to use it. The AI constantly defaults to spamming .clone() or wrapping things in Rc, completely butchering idiomatic Rust and making the output a pain to work with.
This hasn't been true since around gpt-4.5 on the OpenAI side of things. The 5.x models have been pretty much solid on Rust for a while now.
Meh - doubt it. I'm using an iPhone 14, it feels like it works basically the same as it did when I got it. In general slow downs are apps doing and not caring to per tune because the baseline perf expectation of app developers increases to the point where those things don't matter. Use the default apps and you're pretty much ok.
reply